NavCoT: Boosting LLM-Based Vision-and-Language Navigation via Learning Disentangled Reasoning

📄 arXiv: 2403.07376v2 📥 PDF

作者: Bingqian Lin, Yunshuang Nie, Ziming Wei, Jiaqi Chen, Shikui Ma, Jianhua Han, Hang Xu, Xiaojun Chang, Xiaodan Liang

分类: cs.CV, cs.AI, cs.CL, cs.RO

发布日期: 2024-03-12 (更新: 2025-03-22)

备注: Accepted by TPAMI 2025

🔗 代码/项目: GITHUB

💡 一句话要点

提出NavCoT以解决VLN任务中的领域差距问题

🎯 匹配领域: 支柱二：RL算法与架构 (RL & Architecture) 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 视觉与语言导航 大型语言模型 自我引导决策 领域适应 链思维生成

📋 核心要点

  1. 现有方法在视觉与语言导航任务中，常因领域差距导致导航推理准确性不足。
  2. 论文提出的NavCoT通过参数高效的领域内训练，增强了LLM的自我引导导航决策能力。
  3. 实验结果表明，NavCoT在多个VLN基准上表现优异，尤其在R2R数据集上实现了约7%的相对提升。

📝 摘要（中文）

视觉与语言导航（VLN）作为具身人工智能的重要研究问题，要求具身代理根据自然语言指令在复杂的3D环境中导航。近期研究表明，大型语言模型（LLMs）在VLN中的潜力，但由于其离线使用，通常面临VLN任务与LLM训练语料之间的显著领域差距。本文提出了一种新策略——导航链思维（NavCoT），通过参数高效的领域内训练，实现自我引导的导航决策，从而显著减小领域差距。具体而言，在每个时间步，LLM被提示预测导航链思维，依次进行想象、选择和决策。实验结果显示，NavCoT在多个VLN基准上显著优于直接动作预测变体，并在R2R数据集上相较于最新的基于GPT4的方法提高了约7%。

🔬 方法详解

问题定义：本文旨在解决视觉与语言导航任务中，现有大型语言模型因离线训练而导致的领域差距问题。这种差距影响了模型的导航推理准确性和可解释性。

核心思路：NavCoT的核心思路是通过参数高效的领域内训练，使LLM能够自我引导进行导航决策。具体而言，模型在每个时间步通过想象、选择和决策的过程来生成导航链思维。

技术框架：整体架构包括三个主要模块：1) 作为世界模型进行想象，预测下一个观察；2) 选择与想象最一致的候选观察；3) 基于前两步的推理确定行动。

关键创新：NavCoT的主要创新在于通过构建形式化标签进行训练，使LLM能够生成合理的导航链思维输出，从而提升决策能力。这与现有方法的直接动作预测形成了本质区别。

关键设计：在模型设计上，采用了参数高效的微调策略，损失函数设计为能够优化链思维生成的合理性，确保模型在不同训练设置下的适应性和性能。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示，NavCoT在多个流行的VLN基准（如R2R、RxR和R4R）上表现优越，相较于直接动作预测变体，R2R数据集上实现了约7%的相对提升，展示了其显著的性能优势。

🎯 应用场景

该研究的潜在应用领域包括智能机器人、自动驾驶和虚拟助手等具身人工智能系统。通过提升导航决策的准确性和可解释性，NavCoT有助于开发更具适应性和可扩展性的具身代理，推动实际机器人应用的发展。

📄 摘要（原文）

Vision-and-Language Navigation (VLN), as a crucial research problem of Embodied AI, requires an embodied agent to navigate through complex 3D environments following natural language instructions. Recent research has highlighted the promising capacity of large language models (LLMs) in VLN by improving navigational reasoning accuracy and interpretability. However, their predominant use in an offline manner usually suffers from substantial domain gap between the VLN task and the LLM training corpus. This paper introduces a novel strategy called Navigational Chain-of-Thought (NavCoT), where we fulfill parameter-efficient in-domain training to enable self-guided navigational decision, leading to a significant mitigation of the domain gap in a cost-effective manner. Specifically, at each timestep, the LLM is prompted to forecast the navigational chain-of-thought by: 1) acting as a world model to imagine the next observation according to the instruction, 2) selecting the candidate observation that best aligns with the imagination, and 3) determining the action based on the reasoning from the prior steps. Through constructing formalized labels for training, the LLM can learn to generate desired and reasonable chain-of-thought outputs for improving the action decision. Experimental results across various training settings and popular VLN benchmarks (e.g., Room-to-Room (R2R), Room-across-Room (RxR), Room-for-Room (R4R)) show the significant superiority of NavCoT over the direct action prediction variants. Through simple parameter-efficient finetuning, our NavCoT outperforms a recent GPT4-based approach with ~7% relative improvement on the R2R dataset. We believe that NavCoT will help unlock more task-adaptive and scalable LLM-based embodied agents, which are helpful for developing real-world robotics applications. Code is available at https://github.com/expectorlin/NavCoT.